Results 1 to 4 of 4
  1. #1
    Join Date
    Dec 2007
    Posts
    2

    Unanswered: load is slow on a 64 CPU machine than a 2 CPU machine

    Friends

    I recently moved a project to production and found that the load which was running for 3 minutes in a 2 CPU dev box is now taking 6 - 7 minutes in 64 CPU with 62 partitions on production box.This is the first time we are loading to a prodution box for that database. The load is from a remote abinitio server. I checked db2diag.log and found that pre-partioning agent for the load keeps running till the end and moment he pre-paritioning agent finished next moment I have a load and build phase completed on last of the nodes. Can this be a network issue that data is getting transffered slowly, if yes how can i find that out also i have no access to production and will need to provide scripts to production dba's to provide me with information. Till now i looked at db cfg to check util_heap_sz which was 15000 and that seemed adequate to me.I am looking for suggestions and ideas on what should be next steps in my investigation or if i missed anything in my investigation. and any ideas on what i should ask production dbas to monitor during the next load or any other suggestions...

    Also forgot to mention that they are loaidng 3 tables simultaneously but load file is small in the range of 200 MB or so

    Thanks in advance

  2. #2
    Join Date
    May 2003
    Location
    USA
    Posts
    5,737
    Are you using a load specifically designed for DPF, or a normal load utility. Unless your commands or SQL is operating in parallel, then a DPF machine will always take much longer. By parallel, I don't mean multiple loads at one time.
    M. A. Feldman
    IBM Certified DBA on DB2 for Linux, UNIX, and Windows
    IBM Certified DBA on DB2 for z/OS and OS/390

  3. #3
    Join Date
    Dec 2007
    Posts
    2
    yes we are using load commands specific to DPF. I can see notes in db2diag.log that we are starting load phase in all the nodes parallely but a 200 MB load file is taking 6 minutes which is tooooo much time. The way I started the investigating was to look at db cfg parameters and looking at db2diag.log. The db2diag.log shows that the load on last of the node finished immeditaly after the pre-partioning agent finishes. My understanding of load is that pre-partitioning agent is just supposed to read the file and pass it on the partioning agent and the does the fact that pre-partioning agent is running for a long time mean that file is taking long time in network to transfer. How do i proceed with investigation should I look for something else in db2diag.log or some other file. Basically I need to investigate and come up with some conclusions on what has happened so any pointers on how I should proceed with my investigation would be greatly appreciated.

  4. #4
    Join Date
    Jun 2003
    Location
    Toronto, Canada
    Posts
    5,516
    Provided Answers: 1
    Looks like partitioning is the bottleneck. Try specifying ANYORDER and PARTITIONING_DBPARTNUMS options of the LOAD command to increase the number of concurrent partitioning agents.
    ---
    "It does not work" is not a valid problem statement.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •